WCAG 1.0 CP 6.1 Considered Harmful [was: Re: who does what]

[Summary: WCAG 1.0 CP 1.6 is IMO harmful, should be revoked in the errata
for WCAG 1.0, and purged altogether from WCAG 2.0.]

[...]
> In our last discussion of the issue, we felt that the scenarios that
> were being used to make the case fell under the requirement to
> use style sheets with HTML in a way that when style sheets fail,
> are not supported or turn off that the content is still readable.  For
> example, a sidebar should fall into the flow of the content (likely
> at the beginning or the end).

So... the arguement has been made that classes do not need to have
semantics attached to them because of checkpoint 6.1 in WCAG 1.0? Well, I
think that exposes a *serious* flaw in that checkpoint.

There is *no* default rendering for markup documents on the Web that are
not in some way recognized by the UA. For example, if one feeds an XML
document as-is into MSIE5+ or Ns6+, IE5 shows the outline of the document,
and Ns just renders the text as if the markup weren't there. It is only
when one attaches a stylesheet through use of the xml-stylesheet PI that
any presentation is added to the document at all (via. CSS and or XSLT).

When a browser recieves a document that is of text/html MIME type
(according to what it can deduce from the HTTP headers), then it recognizes
that the document should be rendered as text/html, and applies its
knowledge base of HTML to the document for output. That includes behaviours
for form controls etc., but most noticably it includes the presentational
aspect of the document. In other words, there is a default stylesheet
*built-into* the browser.

Now, most browsers will not let one turn off the built-in stylesheet (I
think Ns6+ lets one play with it), but that does not mean that it isn't
there. A default UA presentation is still what WCAG considers to be a
"stylesheet", and yet when it is turned off, a page will often not still be
readable. So all pages are in violation of CP 6.1. That is not the fault of
the Web pages of the world, it is a fault with CP 6.1, and that is the
reason I believe CP 6.1 to be a checkpoint which should be revoked in the
errata for WCAG 1.0, and purged altogether from WCAG 2.0.

The semantics of HTML are well known because of HTML's pervasiveness. But
one should not conclude that default (especially visual) renderings are
inherent to HTML itself. If I want to override <p> displaying as a block
level element, then I can quite easily do so:-

   p { display: inline; }

HTML has no *standard* "view". There are hints in the HTML 4.01
soecification as to how it MAY be rendered, but these barely constitute a
*default* rendering, yet alone a *standard* rendering. XHTML 2.0 will
hopefully break totally clear of these idioms, and I don't expect any XHTML
2.0 UA to have a built in stylesheet. It doesn't really need to have it.

I think Nick Kew put it best in his recent www-validator rant:-

[[[
(1) The novice (or non-technical website owner) question:
    "my site looks right and works fine - isn't that enough?"

The answer to this one is that markup languages are no more than
data formats.  So a website doesn't look like anything at all!
]]] - http://lists.w3.org/Archives/Public/w3c-wai-gl/2001JulSep/1071

So, this brings us back to the question of whether or not classes should
have semantics associated with them. Since I have just discussed how CP 6.1
is in ignorance of some pretty basic Web axioms, this checkpoint cannot be
used as a fallback. Classes are (whether a good mechanism or not) a method
of attaching additional semantics to elements in XHTML, and as such, the
author MUST have a way to make those additional semantics known. Otherwise,
classes are ambiguous, and therefore are not fully accessible.

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://purl.org/net/sbp/> .

Received on Tuesday, 25 September 2001 13:04:02 UTC